AI/Rock-X

    Home >  AI Development  >  ROCK X SDK

ROCK X SDK

Introduction

ROCK-X SDK is a set of AI components libraries for RK3399Pro/RK1808 platform. Developers can build rapid AI applications with ROCK-X SDK API. Currently the following features are provided:

Classification	Functions
Object	Head Detection / 91 Classes Object Detection
Face	Face Landmark / Face Analyze / Face Recognition
Plate	Car Plate Detection / Car plate Recognition
Pose	Body Keypoint / Finger Keypoint

System requirement

On the RK3399Pro platform, the libraries and applications provided by the SDK require RKNN driver version 0.9.6. When running Demo in Android/Linux , you can see the following driver information through log information:

   ==============================================
   RKNNVERSION:
      API: 0.9.5 (a949908 build: 2019-05-0722:20:52)
      DRV: 0.9.6 (c12de8a build: 2019-05-0620:10:17)
   ==============================================

Make sure the DRV version is 0.9.6

On the RK1808 platform, the libraries and applications provided by the SDK require rknn_runtime version above 0.9.6. The rknn_runtime version can be viewed on the RK1808 platform as follows:

   $ strings /usr/lib/librknn_runtime.so | grep "librknn_runtime version"
   librknn_runtime version 0.9.8 (8dcfdc7 build: 2019-05-31 11:23:34 base: 110)

Using SDK

After version 1.1.2, ROCK-X SDK provided C and Python interfaces.

C Interface

1. The sample application

The C interface sample application code includes the command line execution program example and the Android program example. To compile and run, refer to the README file in the demo directory.

2. Import the SDK libraries

Rock-x SDK libraries for each platform are located in the SDK directory, as shown below:

   sdk /
   ├── rockx-data
   ├── rockx-rk1808-Linux
   ├── rockx-rk3399pro-Android
   ├── rockx-rk3399pro-Linux
   └── rockx-x86-64-Linux

Developers only need to introduce the corresponding platform library in their project's CMakeLists.txt.

3. API Document

More detail please refer to: http://t.rock-chips.com/wiki/rockx_api_doc/

Python Interface

1. Install The Wheel Package

The Python wheel package is located in the SDK/Python directory, so developers need to install Python3 and the python3-pip environment first, Then install it directly from the following command:

   pip3 install RockX-*-py3-none-any.whl

The Demo also relies on the opencv-python package to run.

2. The sample application

The Python installation package comes with the Demo of SDK interface, which can be run directly after the installation. Code for each Demo located in the sdk/python/test directory, developers can use for reference development.

3. API Document

The API functions of the RockX python SDK can be viewed through the help function that comes with python, as shown below:

   $ python3
   >>> from rockx import RockX
   >>> help(RockX)